NSD2 Promotes Tumor Angiogenesis Through Methylating and Activating STAT3 Signaling

Background (cid:0) Tumor angiogenesis plays important roles in tumorigenesis and development, the regulation mechanism of angiogenesis is still not been fully elucidated. Nuclear receptor binding SET domain protein 2 (NSD2), a histone methyltransferase which catalyzes the di-methylation of histone H3 at lysine 36, has been proved a critical molecule in proliferation, metastasis and tumorigenesis. But its role in tumor angiogenesis remains unknown. Methods (cid:0) Cell Counting Kit 8 (CCK8), scratch assays, transwell-migration assays, tube-formation and mice xenograft model assays were used to conrm the role of NSD2 in the bioprocess of angiogenesis. Bioinformatics analysis, western blot and immunouorescence staining were used to verify the function of NSD2 in regulating STAT3 signaling pathway. Immunouorescence co-localization, immunoprecipitation, mass spectrometry and site-directed mutagenesis were performed to determine the NSD2-dependent methylation site of STAT3. Results: Here we demonstrated that NSD2 promoted tumor angiogenesis in vitro and in vivo. Furthermore, we conrmed that the angiogenic function of NSD2 was mediated by STAT3. Momentously, we found that NSD2 promoted the methylation of STAT3 and that the inhibition of STAT3 methylation resulted in the attenuation of STAT3 signaling pathway. In addition, mass spectrometry and site-directed mutagenesis assays revealed that NSD2 methylated STAT3 at lysine 163 (K163), and K163 was of signicance in the activation of STAT3 signaling pathway. Conclusion: We conclude that methylation of STAT3 catalyzed by NSD2 promotes the activation of STAT3 pathway and enhances the ability of tumor angiogenesis. Our ndings investigate a NSD2 dependent methylation-phosphorylation regulation pattern of STAT3 and reveal that NSD2/STAT3/VEGFA axis might be a potential target for tumor therapy. NSD2 promotes the STAT3 activation. A and GSEA analysis was performed in two different GEO datasets to reveal the association between the expression of NSD2 and the activation of IL-6/JAK/STAT3 pathway. NSD2 nuclear-location


Background
Angiogenesis is required in both tumor growth and metastasis and is considered as one of the ten hallmarks of cancer [1]. Because of the rapid proliferation signature of cancer, new blood vessels arise from pre-existing microvasculature to meet the demand of tumor progression [2]. Normally, angiogenesis is strictly under the control of angiogenic and anti-angiogenic factors. However, in some conditions such as in ammation, wound healing, and tumorigenesis, the balance is disrupted and blood vessel formation is promoted [3]. Among the most common factors in uencing angiogenesis including VEGF, EGF, HGF, b-FGF, etc, VEGF seems to be the most important molecule [4][5][6]. VEGF antagonists, such as bevacizumab signi cantly attenuate angiogenesis and have been used in the treatment of multiple carcinomas [7,8]. However, similar to most targeted drugs, VEGF antagonists also encounter both primary and secondary resistance, which is associated with the failure of VEGF antagonists [9]. Therefore, an in-depth exploration of the mechanism of tumor angiogenesis as well as its regulation is crucial and may be of great help in guiding the clinical targeted therapy.
Signal Transducers and Activators of Transcription (STATs) are a vital protein family of transcription factors widely studied in cancer research. STAT3 was identi ed at rst as an acute phase response factor activated by the cytokine interleukin-6 (IL-6) [10,11]. STAT3 plays a key role in several biological processes in carcinoma, including angiogenesis, proliferation, resisting apoptosis, evading the immune response, and so on, and nearly all the hallmarks of cancer proposed by Weinberg are under the regulation of STAT3 [12,13]. In recent years, non-classical post-translational modi cations such as methylation, acetylation, ubiquitination, and SUMOylation, has gained considerable attention in recent years [14]. STAT3 is reported to be methylated by SET9 and EZH2 at different residue sites resulting in different transcription activating status of STAT3 [15][16][17][18]. Acetylation of STAT3 catalyzed by acetyltransferase p300 is reported to be essential in the formation of STAT3 homodimer, which is thought to play an important transcription factor function [19].
NSD2 (also known as WHSC1 and MMSET) is a histone methyltransferase catalyzing the demethylation of histone 3 at lysine 36 (H3K36me2) and is associated with active transcription of a series of genes [20]. NSD2 plays a signi cant role in cell development, and NSD2 haploinsu ciency is associated with Wolf-Hirschhorn syndrome (WHS) which is a multiple malformation syndrome [21]. Meanwhile, NSD2 knockedout mice exhibited embryonic development disorder [22]. In recent years, the function of NSD2 in all cancer types have been gradually revealed. Multiple myeloma (MM), one of the most fatal hematologic malignancies, is often characterized by chromosomal translocation [23]. Among them, the t (4; 14) translocation, which is one of the major types of chromosomal translocation in MM, is associated with the overexpression of NSD2 and leads to a poor prognosis [24][25][26]. NSD2 is overexpressed in prostate cancer, especially in metastatic niche and has been proven to be associated with the poor prognosis in prostate cancer patients [26], NSD2 can also directly methylate PTEN and enhance the DNA damage repair ability in colorectal cancer, and as a result, enhance the resistance of cancer cells to chemotherapy [27]. Similar results have been reported in other solid carcinomas such as esophagus carcinoma, stomach carcinoma, hepatocellular carcinoma, lung cancer, corpus uteri malignancy, etc [28]. However, the function of NSD2 in angiogenesis remains to be explored.
Here, we demonstrated NSD2 interacted with STAT3 enhancing its methylation and phosphorylation and resulting in transcription activation of VEGFA and promotion of angiogenesis. Moreover, this study identi ed Lysine 163 (K163) of STAT3 as the NSD2-dependent methylation site, and this residue could be of signi cance in the activation process of STAT3 signaling pathway. The ndings indicate that NSD2 plays a vital role in angiogenesis and could be a potential therapeutic targeted for tumor angiogenesis.
Stable cell lines and plasmid construction Lentiviral vectors containing shNC (negative control, NC) and shRNA-NSD2 were constructed using the vector PLKO.1. The full-length NSD2 with HA-tag was cloned into a lentiviral vector PLKO.AS3w which was used for overexpressing stable cell lines. The full-length STAT3 with Flag-tag was cloned into the pcDNA3,1(+) vector. Full-length STAT3 was divided into three parts, and the three fragment plasmids were constructed. STAT3 K163R and STAT3 K244R were generated using a Mut Express II Fast Mutagenesis Kit V2 (Vazyme, Nanjing, China). All the primers used are listed in Table S1.
Cell Counting Kit-8 (CCK8) assay HUVECs (2.5 × 10 3 /100 μl) were seeded into 96-well plates and treated with CM from altered disposed cell lines. At the indicated times (0 h, 24 h, 48 h and 72 h), 10 μl of CCK was added to the medium. Following incubation for 2 h, the cell proliferation rate was assessed by measuring the absorbance at 450nm of the medium.
Cell migration assay HUVECs (5 × 10 4 /200 μl) were suspended using medium without serum. The cells were then seeded into the upper chamber of a transwell chamber, and 650 μl of CM from different disposed CRC cells was added to the lower chamber. The cells were incubated for 16 h and stained with a violet solution. Images were captured using a light microscope.

Wound healing assay
HUVECs were seeded into 6-well plates. When 98% of cell density was reached, the media were replaced with CM from altered cells including 1μM 5-uouracil to block the proliferation after scratch wound, and a 200 μl sterile plastic pipette tip was used to scratch the cells. The wound distances were measured using a light microscope. Within 24 h, the wound healing distances were photographed and measured.

Tube formation assay
Growth factor-reduced matrigel (BD Biosciences, NJ, USA). was placed into 96-microwell plates 50μL per well and solidi ed at 37 °C for 30 min. Then, 4 × 104 HUVECs were separately suspended in CM from different cells and seeded into the pre-coated matrigel microwells. They were incubated for 8 h afterwhich tube formation was observed using an inverted microscope (Eclipse model TS100; Nikon, Tokyo, Japan).
The tube formation ability of the HUVECs was assessed by counting the number of branch points per eld.
Immuno uorescence assay Cells expressing shNC and shNSD2 were seeded on coverslips. The cells were xed with neutral tissue xator and washed with PBS three times. The xed cells were permeabilized with 0.1% Triton X-100 for 2 min, blocked with 1% bovine serum albumin for 1 h, and treated with the indicated primary antibodies at 4OC overnight. The cells were then treated with secondary antibodies and incubated for 1 h and DAPI for 15 minutes at room temperature. Visualization of the cells was done using a confocal laser scanning microscope (Olympus FLUOVIEW FV1000).

Western blot
Protein samples from tissues and cells were electrophoretically separated by SDS-PAGE and transferred to PVDF membranes (Millipore, MO, USA). Molecular weight speci c bands were incubated with their corresponding primary antibodies at 4 °C overnight and were then incubated with HRP-conjugated secondary antibodies for 2 h at room temperature. The bands were visualized with ECL reagents (Thermo Fisher, MA, USA).

Immunoprecipitation (IP) and quanti ed immunoprecipitation (qIP) assay
Cells were harvested and lysed in NP40 lysis buffer for 30 min. Supernatants were precleared with 10 μl of Protein A/G Magnetic Beads (#HY-K0202, MCE, NJ, USA) for 2 h. For the endogenous IP assay, 3-5 μL antibodies against NSD2 or STAT3 were added into 40 μL magnetic beads that had been resuspended in 500 μL NP40 lysis buffer. The mixture was incubated at 4 ℃ overnight. The magnetic beads were isolated from the mixture, protein supernatants were added to them, and incubated at 4 ℃ for 4 h. For exogenous IP or qIP, Anti-ag or Anti-HA magnetic beads (#B26101 and #26201, Bimake, TX, USA) were used to enrich HA-taged or Flag-taged proteins. The magnetic beads and protein supernatants were incubated for 4 h at 4℃. The beads were then boiled with an SDS-loading buffer. The protein samples were used for the western blot assay.

Quantitative Real-time PCR (qRT-PCR)
Total RNA was extracted using RNAiso (Takara, Japan) according to the manufacturer's instructions. Synthesis of cDNA was done using ABScript II RT Master Mix (RK20403, Abclonal Technology, Wuhan, China). Quanti cation of the mRNA for the indicated genes was done using an ABI 7300 QuantStudio3 PCR (RT-PCR) System using Genious 2X SYBR Green Fast qPCR Mix (RK21206, Abclonal Technology, Wuhan, China). Gene-speci c primer sequences were designed from the PrimerBank database (Table S1) [29].

Mouse tumor xenograft model
Six-week-old male BALB/c nude mice were purchased from Huafukang Bio-Technology (Beijing, China). Eighteen mice were randomly grouped into three groups. The mice were subcutaneously injected with 5 × 10 5 SW480 cells expressing shNC and shNSD2. At 8 days post injection, the tumor volume was examined after every 4 days. After 28 days, mice were sacri ced for subsequent experiments. Similarly, 5 × 10 5 LoVo cells expressing Vector and NSD2 were injected subcutaneously on the mice. At 8 days post injection, the tumor volume was examined every 4 days, meanwhile vehicle or STATTIC was intra tumor injected every two days. After 24 days, mice were sacri ced for subsequent experiments. The animal experiments were performed according to the NIH Guide for the Care and Use of Laboratory Animals with the approval of the Institutional Animal Care and Use Committee at Tongji Hospital.

Modi cation residue identi cation by mass spectrometry detection
The STAT3 protein sample enriched by the IP assay was separated by SDS-PAGE. After Coomassie bright blue staining, the gel band of interest was excised and sent to the National Protein Science Facility, School of Life Science, Tsinghua University. The gel band was digested, dried, and redisolved in 0.1% tri uoroacetic acid. Peptides were analyzed by an Orbitrap Fusion mass spectrometer.
The MS data was searched against the target protein database from UniProt using an in-house Proteome Discoverer (Version PD1.4, Thermo-Fisher Scienti c, USA). The search criteria were: full trypsin speci city was required; carbamidomethylation (C) were set as the xed modi cations; the oxidation (M) and methyl (K/R) were set as the variable modi cation; precursor ion mass tolerances were set at 20 ppm for all MS acquired in an orbitrap mass analyzer; and the fragment ion mass tolerance was set at 0.02 Da for all MS2 spectra acquired.

Statistical analysis
All data were analysed with GraphPad Prism 7.0 (LaJolla, CA, USA). Two-tailed Student's t-tests and ANOVA analysis were used. Statistical signi cance was set at p < 0.05.

Results
Inhibition of NSD2 moderates tumor-induced angiogenesis in vitro NSD2 has been reported to promote proliferation, chemotherapy-resistance, migration, and invasion in altered malignancies. However, until now, little has been revealed about the role of NSD2 in tumor angiogenesis. Since NSD2 is reported to be up-regulated and predicts poor prognosis in varieties of carcinomas, and angiogenesis plays an important role in tumor progression, the relationship between NSD2 and tumor-induced angiogenesis requires to be explored. In this study, we performed a GSEA analysis using publically available datasets and found that the up-regulation of NSD2 was related to the bioprocess of angiogenesis (Fig.1A). To detect the function of NSD2, we constructed stable NSD2 shRNA-NSD2(shNSD2)-expressing SW48 and SW480 cell lines using two lentiviruses containing two altered shRNA sequences against NSD2 (Fig.1B). A CCK8 assay was performed to evaluate the proliferation ability in human umbilical vein endothelial cells (HUVECs) incubated with different condition medium (CM). The proliferative ability of HUVECs was found to be moderate in cells expressing shNSD2 compared with those expressing shRNA-NC (shNC) (Fig.1C and 1D). Besides, the proliferation and migration of HUVECs were found to be vital for angiogenesis. Therefore, a scratch assay was performed to evaluate the migration ability of HUVECs, and the results showed that CM from shNSD2 cells inhibited the migratory ability of HUVECs (Fig. S1A, S1B, S1C and S1D). As expected, in a transwell-migration assay, similar results were obtained showing that CM from shNSD2 cells signi cantly attenuated HUVECs migration compared with CM from shNC cells (Fig.1E, 1F and 1G). A tube formation assay was performed using HUVECS incubated with CM as mentioned above. Less tube formation was observed in the shNSD2 groups compared with the shNC groups (Fig.1H, 1I and 1J). In conclusion, NSD2 attenuates tumor-induced angiogenesis in vitro.

Loss of NSD2 attenuates the expression of VEGFA and inhibits angiogenesis in vivo
Studies reveal that a series of angiogenic factors are involved in tumor-induced angiogenesis. In this study, we showed that NSD2 could promote tumor angiogenesis, therefore we further investigated if NSD2 could in uence the spectrum of angiogenic factors. To verify this hypothesis, we detected the mRNA expression level of several angiogenic factors such as VEGFA, PDGFB, etc. with the intervention of NSD2. Loss of NSD2 was found to inhibit the expression of VEGFA ( Fig.2A). Besides mRNA expression, NSD2 also in uenced the expression of VEGFA protein (Fig.2B). As an angiogenic factor, only secretory VEGFA could promote angiogenesis in the tumor microenvironment (TME). Therefore, it was important to investigate whether NSD2 in uenced the secretion of VEGFA. As expected, the ELISA assay revealed that NSD2 knockdown decreased the levels of secretory VEGFA in CM ( Fig.2C and 2D).
To con rm the function of NSD2 in promoting angiogenesis and upregulating VEGFA, a mouse xenograft model was set up using shNC and shNSD2 SW480 cell lines. After 28 days, mice were sacri ced and the xenografts gathered. Xenografts formed by shNSD2 cells were smaller in appearance compared with those formed by shNC cells (Fig.2E). After tumors became visible to naked eyes, tumor volumes were calculated every 4 days and a tumor growth curve is drawn, which indicated that tumors in the shNSD2 group grew slowly compared with the shNC group (Fig.2F). After the xenografts were collected, we measured the weight of these xenografts and found that xenografts formed using NSD2 downregulated cells were lighter in weight (Fig.2G). Besides, the protein expression of NSD2 and VEGFA of the xenograft was detected using western blot assay, and the results showed that the loss of NSD2 attenuated the expression of VEGFA in vivo (Fig.2H). Consistent with the above results, IHC staining was performed and revealed that knockdown of NSD2 inhibited the expression of VEGFA (Fig3.I). IHC staining results of CD31, which was considered one of the major markers of vascular endothelial cells, revealed that tumor vessel density decreased with NSD2 intervention (Fig.2J). These ndings indicate that loss of NSD2 inhibits angiogenesis by downregulating expression of VEGFA in vivo.
NSD2 in uences the activation of the STAT3 signaling pathway INCB054329, a bromodomain and extraterminal domain (BET) inhibitor, could be regarded as a potent inhibitor of NSD2 and was reported to suppress the JAK-STAT pathway [30]. Meanwhile, GSEA analysis was performed in two different GEO datasets and results showed that the up-regulation of NSD2 was signi cantly associated with the IL-6/STAT3 pathway (Fig.3A and 3B). Therefore, we put forward the hypothesis that NSD2 activates the STAT3 signaling pathway to up-regulate the expression of VEGFA promoting tumor angiogenesis. To test this hypothesis, we detected the phosphorylation and total levels of STAT3 in multiple cell lines. Western blot analysis revealed inhibition of NSD2 could impair the phosphorylation of STAT3, while overexpression of NSD2 led to an increase in STAT3 phosphorylation ( Fig.3C and 3D). STAT3 functions as a transcription factor, hence, the nuclear translocation extent of STAT3 can re ect its functional status to some extent. The subcellular location of STAT3 upon the intervention of NSD2 was detected, and results showed that NSD2 knockdown resulted in the translocation of STAT3 from the nucleus to the cytosol (Fig.3E). Conversely, the overexpression of NSD2 promoted STAT3 translocation from the cytosol to the nucleus (Fig.3F). Consistently, an immuno uorescence assay was performed, and as expected, when the expression of NSD2 was attenuated, the aggregation of STAT in the nucleus also decreased (Fig.3G). Taken together, NSD2 in uenced the activation of the STAT3 signaling pathway.
The angiogenic function of NSD2 is mediated by STAT3 signaling pathway STAT3 signaling pathway plays a vital role in the regulation of VEGFA secretion and angiogenesis under physiological and pathological conditions. In this study, we proved that NSD2 promoted angiogenesis, VEGFA expression, and upregulation of the STAT3 signaling pathway. We then determined if the function of NSD2 was dependent on the STAT3 signaling pathway. NSD2 was overexpressed with or without STAT3 knockdown and the phosphorylation of STAT3 and expression of VEGFA detected by western blot assay. Increased phosphorylation of STAT3 and upregulation of VEGFA were reversed by STAT3 knockdown (Fig.4A and 4B). CCK8 assay was performed to evaluate the proliferative ability of HUVECs incubated with CM from tumor cells. The Result showed that HUVECs incubated with CM from NSD2 overexpressing cells grew faster, but these results were reversed by STAT3 intervention (Fig.4C and 4D). The migration ability of HUVECs incubated with altered CM was measured by stretch and transwell assay. Consistently, HUVECs incubated with CM from NSD2 overexpressing cells exhibited better migratory ability and this effect was inhibited by STAT3 intervention (Fig.4E, 4F, 4G, S2A, S2B, S2C and S2D). Furthermore, the angiogenic effect of CM from cells expressing NSD2 was enhanced as detected by HUVECs tube-formation assay, while the angiogenic function of NSD2 was reversed by attenuating STAT3 expression (Fig.4H, 4I and 4J). Consistent with the above results, ELISA results showed that the concentration of VEGFA in CM was elevated in NSD2 overexpressing cells but was reversed by downregulating STAT3 expression (Fig.S2E and S2F). Taken together, the angiogenic function of NSD2 was reversed by STAT3 intervention. Therefore, we concluded that the angiogenic function of NSD2 is mediated by STAT3 signaling pathway.

STAT3 inhibitor STATTIC abolishes the angiogenic function of NSD2 in vivo.
The transcription factor STAT3 plays an important role in oncogenesis and tumor development, hence, targeted drugs against the STAT3 signaling pathway have been widely studied. Early-stage clinical trials have been conducted using STAT3 inhibitors in altered tumors. Here, STATTIC, which is considered an effective STAT3 inhibitor was selected to observe the therapeutic effect in NSD2-overexpressing carcinomas. The effect of STATTIC in the LoVo cell line overexpressing NSD2 was investigated. Surprisingly, NSD2 was found to upregulate the phosphorylation of STAT3 and the expression of VEGFA, and these effects were reversed by the addition of STATTIC (Fig.5A). Next, a xenograft model was set up using LoVo cell line to observe the effect of STATTIC in the NSD2 overexpressing tumor cells xenograft. After 24 days, mice were sacri ced and the xenografts were collected and the tumor volume and weight determined. Xenografts formed by NSD2 overexpressing cells were bigger and the effects were reversed by STATTIC (Fig. 5B). The tumor volume was measured in vivo, and a growth curve was drawn to determine the growth status of the xenograft, which showed a consistent result with the above results ( Fig. 5C). Determination of the weight of xenografts revealed that xenografts formed by NSD2 overexpressing cells were heavier and this effect was reversed by the addition of STATTIC (Fig.5D). To verify this hypothesis in vivo, the xenograft homogenate was obtained and a western blot assay was performed. As shown in Fig. 5E, the results were consistent with the in vitro results. IHC assay was performed, and the results showed that phosphorylation of STAT3, expression of VEGFA and CD31positive cell counts were elevated when NSD2 was overexpressed, however, these were reversed by STATTIC (Fig. 5F). According to CD31 staining status, microvessel density was elevated with NSD2 overexpression and was blocked with the addition of STATTIC (Fig. 5G). Taken together, we concluded that NSD2 promoted tumor proliferation and angiogenesis in vivo while these effects were blocked by the use of STAT3 inhibitor STATTIC.

NSD2 directly interacts with and methylates STAT3 to activate the STAT3 signaling pathway
To further investigate the mechanism by which NSD2 regulates the phosphorylation of STAT3, we performed an immunoprecipitation assay in the SW480 cell line and explored the physical interaction between NSD2 and STAT3 endogenously (Fig.6A). Furthermore, we constructed the overexpressing plasmid HA-NSD2 and Flag-STAT3, and both were co-transfected in the HEK293T cell line. An immunoprecipitation assay was performed to verify the exogenous interaction, and a consistent result was obtained (Fig. 6B). To further support this result, immuno uorescence was performed to observe the co-localization status of NSD2 and STAT3 (Fig. 6C). NSD2 is a histone methyltransferase, hence we put forward our hypothesis that NSD2 could methylate STAT3 protein. To con rm this hypothesis, immunoprecipitation was carried out to detect the methylation status of STAT3 with or without NSD2 intervention. We concluded that downregulation of NSD2 resulted in the loss of methylation of STAT3 protein, while overexpression of NSD2 was associated with the hyper-methylation of STAT3 protein ( Fig.  6D and 6E). We further tried to explain the relationship between STAT3 methylation and phosphorylation. DZNep, is an inhibitor reported to be global histone methylation rather than an EZH2 inhibitor, was used to con rm the in uence of STAT3 methylation upon phosphorylation. Surprisingly, we found that NSD2 upregulated both the phosphorylation and methylation of STAT3, while the addition of DZNep successfully blocked the methylation of STAT3 protein and resulted in a decrease in STAT3 phosphorylation (Fig. 6F). To nd out the speci c binding domain between NSD2 and STAT3 and the NSD2 dependent methylation residue of STAT3, we constructed three fragment plasmids of STAT3 based on the natural structural domain (Fig. 6G). Next, we overexpressed these fragment plasmids in the HEK293T cell line and performed an immunoprecipitation assay to determine which fragment of STAT3 combined with NSD2. The results showed that Fragment1 (AA1-320) of STAT3 interacted with NSD2 (Fig.6H). These nds indicated that NSD2 interacted with and methylated STAT3 to activate the STAT3 signaling pathway.
Lysine 163 of STAT3 is vital in NSD2 mediating STAT3 methylation and activation In this study, we just proved that NSD2 directly interacted with and methylated STAT3. To nd out the NSD2-dependent STAT3 methylation site, immunoprecipitation was used to enrich the STAT3 protein for mass spectrometry. Coomassie blue staining and mass spectrometry results revealed that we successfully gathered enough speci c STAT3 protein (Fig. 7A). We identi ed 14 lysine residues in our spectrometry result (Table S2).
Among them, some residues such as K49 and K140 which were already reported to be methylated residues were also observed in mass spectrometry results. Enzyme catalyzed reactions need the interaction between enzymes and substrates. One speci c lysine is catalyzed by one speci c methyltransferase. Therefore, by analyzing the mass spectrometry result and searching for the lysing residues located in the rst fragment of the STAT3 protein and excluding the lysine residues which were already reported to be catalyzed by other enzymes, we have found two possible lysine residues, K163 and K244 as NSD2-dependent methylation sites (Fig. 7B). A secondary mass spectrometry result is shown to prove the K163 methylation status of STAT3 (Fig. 7C). Next, we investigated the target site of the NSD2dependent STAT3 methylation residue between the two possible sites. We constructed K to R mutant plasmid at K163 and K244 respectively, by overexpressing or downregulating NSD2, and immunoprecipitation used to detect the methylation status of STAT3. The results showed that at rst both site mutations attenuated the methylation level of STAT3. The changes in NSD2 expressing level in uenced the methylation status of K244R mutant rather than K163R mutant (Fig. 7D and 7E). These results implied that NSD2 in uenced the methylation of STAT3 at lysine 163. We also analyzed the conservative property between different species and concluded that 163 of STAT3 is conservative in most species (Fig. S3A). Finally, we performed an IP assay and western blot assay to con rm whether K163 was an important residue for the activation of STAT3, by comparing STAT3 wildtype with STAT3 K163R mutant. The results showed that K163 played an important role in the activation of the STAT3 signaling pathway (Fig. 7F). In conclusion, we proved that NSD2 directly interacted with STAT3 and methylated the latter at K163 residue which played a vital role in the activation of the STAT3 signaling pathway (Fig. 7G).

Discussion
In this study, we con rmed that NSD2 is associated with STAT3 to methylate and change its phosphorylation level. Moreover, we identi ed K163 as an NSD2-dependent methylation site of STAT3, which is also responsible for the IL-6 mediated STAT3 phosphorylation. Both EZH2 and SET9 are reported to have the ability to methylate STAT3 at different amino residues upon IL-6 stimulation [15,16]. In this study, NSD2 was found to also take part in this process. Therefore, there is a possibility that not only one methyltransferase or one speci c methylation residue is fully responsible for the activation process. This kind of methylation-phosphorylation crosstalk is a common phenomenon in the regulation of all types of signaling pathways, which guarantees the precise regulation of various biological behaviors. In this study, We con rmed another regulatory role of the STAT3 pathway following the methylation-phosphorylation.
Post-translational modi cation is of great signi cance in regulating protein activities. As one of the posttranslational modi cations, methylation has been greatly researched in recent years. Besides research on histones and non-histone methylation has received considerable attention among researchers. The methylation of non-histone protein may be involved in the regulation of the signaling pathway by in uencing protein stability, affecting the activities of transcription factors, in uencing other posttranslational modi cations, altering the protein-protein interaction (PPI) and so on [31][32][33]. For example, AKT was reported to be methylated by SETDB1 at K64 and K140/142 respectively promoting the activity of PI3K/AKT signaling pathway [34,35]. Even more important, mithramycin, an anti-neoplastic antibiotic is reported to effectively inhibit SETDB1 expression and function, and overcome the resistance of KRAS mutant CRC cells to cetuximab both in vitro and in vivo [36]. NSD2 is a histone methyltransferase responsible for di-methylation of histone3 at lysine 36 (H3K36me2). NSD2 possesses the capacity to methylate histones, hence could have the ability to methylate non-histone proteins as well. A previous study elucidated that NSD2 methylate PTEN at K349, promoting its phosphorylation by ATM under the DNA double-strand breaks condition and promotes DNA damage repair [27]. In another study, NSD2 was found to methylate Aurora kinase A (AURKA) at K14 and K117 with changes in its kinase activity mediating cell proliferation via the p53 signal pathway [37]. In our study, we identi ed STAT3 as an original non-histone protein substrate of NSD2 and elucidated the function of a speci c methylation amino residue. These ndings provided theoretical basis for the usage of potential NSD2 inhibitor and the exploitation of inhibitory peptide to block the interaction between NSD2 and STAT3 and the methylation of speci c residue.
However, the role of NSD2 in tumor angiogenesis is in distinct. In this study, loss of NSD2 was found to impair colon cancer angiogenesis both in vitro and in vivo. Angiogenesis is a process controlled by angiogenic and anti-angiogenic factors [3]. Here, we revealed that VEGFA rather than EGF, FGF, HGF, and other factors took part in the NSD2 mediated tumor angiogenesis.
In conclusion, this study has elucidated the function of NSD2 in regulating VEGFA mediated tumor angiogenesis. Mechanistically, we have identi ed an original methylation amino residue of STAT3 which is dependent on NSD2. Besides, the non-histone protein methylation function of NSD2 in regulating STAT3 signal pathway transduction is reported. These ndings provide evidence that NSD2 antagonists or NSD2-STAT3 interfering peptides can be used as therapeutic targets in targeting angiogenesis.

Conclusion
In summary, our results indicated a methylation-phosphorylation interaction in the process of STAT3 activation, which was mediated by histone methyltransferase NSD2. NSD2 methylates STAT3 on K163, which promotes phosphorylation and nuclear translocation of STAT3, as well as promotes the expression of VEGFA and tumor angiogenesis. Therefore, both NSD2 and K163 methylation of STAT3 could be a potential target for cancer therapy.   The angiogenic function of NSD2 is mediated by STAT3. A and B. Western blot evaluating whether STAT3 was dependent in the NSD2 mediated VEGFA expression promotion by overexpressing NSD2 and silencing STAT3 at the same time. C and D. CCK8 assay assessing the proliferative ability of HUVECs incubated with CMs from cells overexpressed NSD2 and down-regulated STAT3 at the same time. E, F and G. Transwell-migration assay of HUVECs incubated with identi ed CMs was performed to assess the Page 22/24 migratory ability of HUVECs. H, I and J. Tube-formation assay of HUVECs incubated with identi ed CMs re ecting the angiogenic function. Results are presented as mean ± SD for n=3, **p <0.01, ***p <0.001, ****p<0.0001.

Figure 5
STAT3 inhibitor STATTIC abolishes angiogenic function of NSD2 in vivo. A. The effect of STATTIC (10μM, 48 h) was veri ed in LoVo cells with or without NSD2 overexpression by western blot; B. Xenograft model was set up using vesicle or STATTIC (3.75 mg/kg, every two days) by intratumoral injection after xenografts were touchable, the xenografts were collected ather the mice had been sacri ced; C. Tumor volumes were measured after every 4 days and a growth curve drawn; D. After xenograft collection, the tumor weight was measured; E. Tissue homogenates were obtained from these xenografts and a western blot assay was performed; F and G. IHC staining of xenografts using antibodies against NSD2, p-STAT3, VEGFA and CD31, according to the staining status of CD31, MVD was statistically analyzed. NSD2 directly interacts with and methylates STAT3 to activate the STAT3 signaling pathway A. In SW480 cell line, IP assays using antibodies against NSD2 and STAT3 respectively were performed; B. In HEK293T cell line, after co-transfection of HA-NSD2 and Flag-STAT3, IP assays were performed to enrich HA-NSD2 and Flag-STAT3. C. Immuno uorescence assay was performed to observe the co-localization conditions of STAT3 and NSD2 in SW480 cell line; D and E. Quanti ed immunoprecipitation assays and western blot to detect the methylation levels of STAT3 with NSD2 upregulation or attenuation; F. Using global methylation inhibitor DZNep ,and a qIP assay was performed to elucidate the methylation and phosphorylation status of STAT3 with NSD2 overexpression and the relationship between these two modi cations; G. According to the natural structure of STAT3; the full-length STAT3 was divided into three parts and three fragment plasmids were constructed; H. Using three Flag-tagged fragment plasmids of STAT3 and a IP assay was performed to clarify which domain of STAT3 could bind to NSD2.

Figure 7
Lysine 163 of STAT3 is vital in NSD2 mediating STAT3 methylation and activation A. Using IP assay for the enrichment of STAT3 protein, staining with Coomassie bright blue and veri cation by mass spectrometry. B. Screening strategy for predicting the possible methylation residues of STAT3 that is dependent on NSD2 function; C. Secondary mass spectrometry result of one possible methylation residue; D and E, qIP assays were performed by detecting the methylation changes of wild type STAT3, K163R and K244R mutants, respectively, with NSD2 overexpression or intervention; F. A siRNA targeting the 3'UTR was used to knockdown endogenous STAT3 and rescued with wild type STAT3 and K163 mutant, after the stimulation of IL-6, the methylation and phosphorylation of STAT3 were detected; G.
Schematic diagram of our hypothesis about this project.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.